CN106203498B - Urban scene garbage detection method and system - Google Patents
Urban scene garbage detection method and system Download PDFInfo
- Publication number
- CN106203498B CN106203498B CN201610529468.9A CN201610529468A CN106203498B CN 106203498 B CN106203498 B CN 106203498B CN 201610529468 A CN201610529468 A CN 201610529468A CN 106203498 B CN106203498 B CN 106203498B
- Authority
- CN
- China
- Prior art keywords
- urban
- garbage
- deep learning
- detection
- data set
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/251—Fusion techniques of input or preprocessed data
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a method and a system for detecting urban scene garbage. The urban scene garbage detection method comprises the following steps: the method comprises the steps of selecting a visual object classification VOC data set as a basic data set for garbage detection, obtaining a city image, marking out a garbage area, fusing and expanding the city image with the VOC data set and enriching the VOC data set, then building a deep learning platform based on a deep learning technology, selecting a pre-training model on the deep learning platform, carrying out garbage detection on the newly obtained city image through the deep learning platform and the pre-training model after carrying out prior parameter setting on the pre-training model, automatically giving a detection result, not needing a specially-assigned person to take a vehicle for photographing registration and manual garbage area detection, monitoring and detecting all-weather urban unordered garbage discarding conditions, and is low in cost and short in time consumption, so that the method is greatly convenient for detecting and cleaning the unordered garbage in the city, and guarantees urban sanitation and image.
Description
Technical Field
The invention relates to a detection method and a detection system, in particular to a method and a system for detecting urban scene garbage.
Background
The disordered discarded garbage in the city seriously influences the appearance of the city, pollutes the living environment and brings great influence to the city and residents.
In order to clean up the garbage discarded in an unordered manner in a city and maintain the urban sanitation and image, the garbage in an urban scene needs to be detected and positioned, and then the garbage is cleaned according to the positioning. The existing urban scene garbage detection method mainly comprises the steps of dispatching a specially-assigned person to carry out patrol and photographing registration, manually positioning garbage discarded in disorder in the patrol process, operating a handheld camera to photograph, and carrying out arrangement and filing after the patrol to record garbage distribution conditions and corresponding responsible persons. The method needs a specially-assigned person to take a vehicle for photographing and registering, is greatly influenced by the aspects of traffic, weather, personnel vacation, working time and the like, cannot monitor and detect the all-weather urban disordered discarded garbage condition, has the problems of high cost, long time consumption and the like in manual photographing and sorting, is greatly not beneficial to the detection and the cleaning of the disordered discarded garbage in the city, and cannot ensure the urban sanitation and image.
Disclosure of Invention
In view of the above, it is necessary to provide a method and a system for detecting garbage in an urban scene, aiming at the problems that the garbage detection and monitoring in the urban scene cannot be performed all day long and the cost is high and the time consumption is long.
The invention provides a method for detecting urban scene garbage, which comprises the following steps:
s10: selecting a VOC data set as a basic data set for garbage detection, collecting urban images, selecting the urban images containing disordered discarded garbage as a candidate set, labeling areas of the urban images containing disordered discarded garbage in the candidate set according to a format defined by the VOC data set, and fusing the labeled urban images with existing data in the VOC data set;
s20: building a deep learning platform for garbage detection on the basis of the fused VOC data set, acquiring a pre-training model for garbage detection provided by the deep learning platform on the built deep learning platform, and carrying out adaptive prior parameter setting on the pre-training model;
s30: and performing garbage detection on the newly acquired urban image by adopting a pre-training model, detecting whether garbage exists in the newly acquired urban image and a garbage existing area, and giving a detection result.
In one embodiment, the step S10 specifically includes: collecting urban images, selecting the urban images containing disordered discarded garbage as a candidate set, marking the areas containing disordered discarded garbage of the urban images in the candidate set by adopting a rectangular selection frame according to a format defined by a VOC data set, randomly dividing the marked urban images into a training set, a verification set and a test set after marking is finished, and respectively fusing the newly acquired training set, verification set and test set with the existing training set, verification set and test set in the VOC data set.
In one embodiment, the step S20 specifically includes:
and selecting a Caffe deep learning framework to realize a deep learning platform, and using a ZF Model in the Model zoom as a pre-training Model of the garbage detection task.
In one embodiment, the step S20 specifically includes: and verifying the detection precision of different prior parameters on the urban image on the fused VOC data set by using a grid searching method.
In one embodiment, the step S30 specifically includes:
preprocessing the newly acquired urban image, specifically comprising cutting and scaling the newly acquired urban image and/or carrying out mean value extraction processing;
and inputting the preprocessed urban image into a deep learning neural network to obtain the classification and position regression of the candidate regions in the urban image so as to obtain a detection result.
The invention provides a system for detecting urban scene garbage, which comprises:
the data fusion module is used for selecting a VOC data set as a basic data set for garbage detection, collecting urban images, selecting the urban images containing disordered discarded garbage as a candidate set, labeling the areas of the urban images containing disordered discarded garbage in the candidate set according to a format defined by the VOC data set, and fusing the labeled urban images with existing data in the VOC data set;
the deep learning platform building module is used for building a deep learning platform for garbage detection on the basis of the fused VOC data set, acquiring a pre-training model for garbage detection provided by the deep learning platform on the built deep learning platform, and carrying out adaptive prior parameter setting on the pre-training model;
and the urban image garbage detection module is used for detecting garbage of the newly acquired urban image by adopting a pre-training model, detecting whether garbage exists in the newly acquired urban image and detecting a garbage existing area, and giving a detection result.
In one embodiment, the data fusion module collects the urban images, selects the urban images containing disordered discarded garbage as a candidate set, labels the areas of the urban images containing disordered discarded garbage in the candidate set by using a rectangular selection frame according to a format defined by the VOC data set, randomly divides the labeled urban images into a training set, a verification set and a test set after the labeling is completed, and respectively fuses the newly acquired training set, verification set and test set with the existing training set, verification set and test set in the VOC data set.
In one embodiment, the deep learning platform building module selects a Caffe deep learning framework to realize a deep learning platform, and uses a ZF Model in a Model Zoo as a pre-training Model of a garbage detection task.
In one embodiment, the deep learning platform building module verifies the detection accuracy of different prior parameters on the urban image on the fused VOC data set by using a grid search method.
In one embodiment, the urban image garbage detection module performs preprocessing on a newly acquired urban image, specifically includes clipping, scaling, and/or performing mean value extraction on the newly acquired urban image; and inputting the preprocessed urban image into a deep learning neural network to obtain the classification and position regression of the candidate regions in the urban image so as to obtain a detection result. The urban scene garbage detection method and the system select the visual object classification VOC data set as the basic data set for garbage detection, acquire the garbage marked area of the urban image, perform fusion expansion and enrichment of the VOC data set with the VOC data set, then build a deep learning platform based on the deep learning technology, the pre-training model is selected on the deep learning platform, the newly acquired urban image is detected for rubbish through the deep learning platform and the pre-training model after the pre-training model is subjected to prior parameter setting, the detection result is automatically given, no special person is required to take a vehicle for photographing registration and detection of artificial rubbish areas, all-weather monitoring and detection of urban disordered rubbish discarding conditions can be realized, the cost is low, the time consumption is short, therefore, the detection and the cleaning of the disordered discarded garbage in the city are greatly facilitated, and the urban sanitation and image are guaranteed.
Drawings
FIG. 1 is a flow diagram of a city scene spam detection method in one embodiment;
FIG. 2 is a block diagram of a city scene spam detection system in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Fig. 1 is a flowchart of a method for detecting urban scene spam in one embodiment, as shown in fig. 1, the method includes the following steps:
s10: selecting a VOC (visual object classes) data set as a basic data set for garbage detection, collecting urban images, selecting the urban images containing disordered discarded garbage as a candidate set, labeling the areas of the urban images containing disordered discarded garbage in the candidate set according to a format defined by the VOC data set, and fusing the labeled urban images with existing data in the VOC data set.
The VOC (visual object classes) data set is an authoritative scene detection data set, comprises a large number of training verification pictures and test pictures, has a large number of types of marked objects, comprises common objects in urban scenes such as pedestrians, bicycles, buses, cars and motorcycles, and can detect and identify the objects in the urban images through the marked objects in the VOC data set. Therefore, in this embodiment, a VOC (visual object classes) dataset is selected as a basic dataset for garbage detection, and the detection of the garbage discarded out of order in the urban scene is performed by using the data in the VOC dataset and the labeled object.
Due to limited data in the VOC data set, there may be insufficient training data, and in this embodiment, the urban image is acquired to expand the VOC data set. Specifically, the city images are collected, and the city images include city images shot by street view cars, city images crawled by the internet, and the like. Then, the urban images containing the disordered discarded garbage are selected as a candidate set, and areas containing the disordered discarded garbage of the urban images in the candidate set are labeled by adopting a rectangular selection frame according to a format defined by the VOC data set (in the labeling process, the interested objects are completely labeled under the condition of labeling little background as possible). After marking, randomly dividing the marked urban image into a training set, a verification set and a test set, and respectively fusing the newly acquired training set, verification set and test set with the existing training set, verification set and test set in the VOC data set.
Further, the VOC2007 dataset was selected as the basis dataset for spam detection. The VOC2007 data set comprises 5011 training verification pictures and 4952 test pictures, and 20 types of labeled objects are shared and are suitable for being used as a basic data set.
S20: on the basis of the fused VOC data set, a deep learning platform for garbage detection is built, a pre-training model for garbage detection provided by the deep learning platform is obtained on the built deep learning platform, and adaptive prior parameter setting is carried out on the pre-training model.
After the acquired urban image and the VOC data set are fused, on the basis of the fused data, the method detects the disordered discarded garbage in the urban scene based on the deep learning algorithm, builds a deep learning platform for garbage detection, applies the deep learning to the garbage detection of the urban scene, and expands the application of the deep learning. After the deep learning platform for garbage detection is built, the deep learning platform is provided with a plurality of pre-training models, and then the pre-training models suitable for garbage detection are selected and obtained from the plurality of pre-training models. In order to enable the obtained pre-training model to be well suitable for garbage detection in the current region, adaptive prior parameter setting needs to be carried out on the pre-training model.
In order to improve the detection precision, in a specific mode, a Caffe deep learning framework is selected in the step to realize a deep learning platform. The Caffe deep learning framework has complete documents, active communities and rich model libraries, and is suitable for platform building. Meanwhile, the platform hardware configuration adopts GPU (stronger floating point arithmetic capability) as an arithmetic core. Further, Nividia Geforce GTX 980 was used as the GPU, and inter Core i7 and 16G memories were used as the main hardware configuration.
After a Caffe deep learning framework is selected to realize a deep learning platform, the community ecology completed by the Caffe deep learning framework provides rich and well pre-trained models.
In the embodiment, a grid search method is used to verify the detection precision of the urban image by using different prior parameters on the fused VOC data set aiming at the application scene of garbage detection, because different application scenes need to perform different adjustments on the prior parameters of the pre-training model. After repeated verification, the embodiment finally selects 0.001 as the initial learning rate, 0.0005 as the weight attenuation amount and 0.9 as the impulse, and selects 128 candidate regions from each urban image as mini-batch (subset), and performs back propagation of loss to update the pre-training model weight.
S30: and performing garbage detection on the newly acquired urban image by adopting a pre-training model, detecting whether garbage exists in the newly acquired urban image and a garbage existing area, and giving a detection result.
After a deep learning platform is built and a pre-training model is selected to be obtained, the obtained pre-training model is adopted to detect the garbage discarded out of order in the urban scene. Specifically, a city image of a city scene is newly acquired, and then a pre-training model is adopted to perform garbage detection on the newly acquired city image. Whether garbage exists in the urban image or not and whether garbage exists in the area are judged through detection of the pre-training model, and a detection result is given, so that detection of disordered garbage discarding in the urban scene is achieved through deep learning and the pre-training model.
In a further mode, in order to make the urban image meet the requirements of the pre-training model, the newly acquired urban image is preprocessed and cut or scaled, so that the newly acquired urban image meets the requirements of the pre-training model.
Meanwhile, since the pixel range of the urban image is fixed, the mean value extraction process is performed on the newly acquired urban image. Specifically, the pixel values of the red, green and blue three bands in the training set of the fused VOC data set are counted to obtain the mean values of the pixels of the three bands, and the corresponding mean values are subtracted from the three bands of the newly acquired urban image. After preprocessing, inputting the preprocessed urban image into a deep learning neural network, and obtaining the classification and position regression of candidate areas in the urban image to obtain a detection result. And selecting different confidence degrees to screen candidate results according to different application scenes, finding out the most probable region of the disordered discarded garbage, and realizing the detection of the disordered discarded garbage in the urban image.
The urban scene garbage detection method includes the steps of selecting a visual object classification VOC data set as a basic data set for garbage detection, obtaining a garbage region marked by an urban image, fusing and expanding the garbage region with the VOC data set and enriching the VOC data set, then building a deep learning platform based on a deep learning technology, selecting a pre-training model on the deep learning platform, carrying out garbage detection on the newly obtained urban image through the deep learning platform and the pre-training model after the pre-training model is subjected to prior parameter setting, automatically giving a detection result, carrying out photographing registration and manual garbage region detection without a specially-assigned person taking a vehicle, monitoring and detecting all-weather urban disordered discarded garbage conditions, and is low in cost and short in time consumption, so that detection and cleaning of disordered discarded garbage in cities are greatly facilitated, and urban sanitation and image are guaranteed.
Meanwhile, the present invention also provides a system for detecting urban scene garbage, as shown in fig. 2, the system for detecting urban scene garbage comprises:
the data fusion module 100 selects a VOC (visual object classes) data set as a basic data set for garbage detection, collects the urban images and selects the urban images containing unordered discarded garbage as a candidate set, labels the areas of the urban images containing unordered discarded garbage in the candidate set according to a format defined by the VOC data set, and fuses the labeled urban images with existing data in the VOC data set.
The VOC (visual object classes) data set is an authoritative scene detection data set, comprises a large number of training verification pictures and test pictures, has a large number of types of marked objects, comprises common objects in urban scenes such as pedestrians, bicycles, buses, cars and motorcycles, and can detect and identify the objects in the urban images through the marked objects in the VOC data set. Therefore, in this embodiment, the data fusion module 100 selects a VOC (visual object classes) data set as a basic data set for garbage detection, and performs detection of garbage discarded out of order in the urban scene by using data in the VOC data set and labeled objects.
Due to the limited data in the VOC data set, there may be a problem of insufficient training data, and in this embodiment, the data fusion module 100 acquires the urban image to expand the VOC data set. Specifically, the data fusion module 100 collects city images, including city images taken by street view cars, city images crawled by the internet, and the like; then, selecting the urban images containing disordered discarded garbage as a candidate set, and marking the areas containing disordered discarded garbage of the urban images in the candidate set by adopting a rectangular selection frame according to a format defined by the VOC data set (marking the interested objects completely under the condition of marking the background as little as possible in the marking process); after marking, randomly dividing the marked urban image into a training set, a verification set and a test set, and respectively fusing the training set, the verification set and the test set of the newly acquired data with the existing training set, the verification set and the test set in the VOC data set.
Further, the data fusion module 100 selects the VOC2007 data set as the basis data set for spam detection. The VOC2007 data set comprises 5011 training verification pictures and 4952 test pictures, and 20 types of labeled objects are shared and are suitable for being used as a basic data set.
The deep learning platform building module 200 builds a deep learning platform for garbage detection on the basis of the fused VOC data set, acquires a pre-training model for garbage detection provided by the deep learning platform on the built deep learning platform, and sets adaptive prior parameters of the pre-training model.
After the acquired urban image and the VOC data set are fused, on the basis of the fused data, the system detects the disordered discarded garbage in the urban scene based on a deep learning algorithm, the deep learning platform building module 200 builds a deep learning platform for garbage detection, the deep learning is applied to the garbage detection of the urban scene, and the application of the deep learning is expanded. After the deep learning platform for garbage detection is built, the deep learning platform is provided with a plurality of pre-training models, and the deep learning platform building module 200 selects and obtains the pre-training model suitable for garbage detection from the plurality of pre-training models. In order to enable the obtained pre-training model to be well suitable for garbage detection in the current region, adaptive prior parameter setting needs to be carried out on the pre-training model.
In order to improve the detection precision, the deep learning platform building module 200 selects a Caffe deep learning framework to realize the deep learning platform. The Caffe deep learning framework has complete documents, active communities and rich model libraries, and is suitable for platform building. Meanwhile, the deep learning platform hardware configuration adopts GPU (the floating point arithmetic capability is stronger) as an arithmetic core. Further, Nividia Geforce GTX 980 was used as the GPU, and inter Core i7 and 16G memories were used as the main hardware configuration.
After a Caffe deep learning framework is selected to realize a deep learning platform, the community ecology completed by the Caffe deep learning framework provides rich and well pre-trained models, and in the embodiment, the deep learning platform building module 200 uses a ZF Model in a Model Zoo as a pre-training Model of a garbage detection task.
In the embodiment, the deep learning platform building module 200 verifies the detection accuracy of the city image by using different prior parameters on the fused VOC data set by using a grid search method for the garbage detection application scenario because different application scenarios require different adjustments of the prior parameters of the pre-training model. After repeated verification, the embodiment finally selects 0.001 as the initial learning rate, 0.0005 as the weight attenuation amount and 0.9 as the impulse, and selects 128 candidate regions from each urban image as mini-batch (subset), and performs back propagation of loss to update the pre-training model weight.
The urban image garbage detection module 300 performs garbage detection on the newly acquired urban image by using the pre-training model, detects whether garbage exists in the newly acquired urban image and a garbage existing area, and provides a detection result.
After a deep learning platform is built and a pre-training model is selected to be obtained, the urban image garbage detection module 300 detects garbage discarded out of order in an urban scene by using the obtained pre-training model. Specifically, a city image of a city scene is newly acquired, and then a pre-training model is adopted to perform garbage detection on the newly acquired city image. Whether garbage exists in the urban image or not and whether garbage exists in the area are judged through detection of the pre-training model, and a detection result is given, so that detection of disordered garbage discarding in the urban scene is achieved through deep learning and the pre-training model.
In a further manner, in order to make the urban image meet the requirements of the pre-training model, the urban image garbage detection module 300 preprocesses the newly acquired urban image, and cuts or scales the newly acquired urban image, so that the newly acquired urban image meets the requirements of the pre-training model.
Meanwhile, since the pixel range of the urban image is fixed, the urban image garbage detection module 300 further performs mean extraction preprocessing on the newly acquired urban image. Specifically, the urban image garbage detection module 300 counts the pixel values of the three bands of red, green and blue in the training set of the fused VOC data set to obtain the mean values of the pixels of the three bands, and subtracts the corresponding mean values from the three bands of the newly acquired urban image. After preprocessing, inputting the preprocessed urban image into a deep learning neural network, and obtaining the classification and position regression of candidate areas in the urban image to obtain a detection result. And selecting different confidence degrees to screen candidate results according to different application scenes, finding out the most probable region of the disordered discarded garbage, and realizing the detection of the disordered discarded garbage in the urban image.
This urban scene rubbish detecting system, select vision object classification VOC data set as the basic data set that rubbish detected, obtain and merge the extension and richen VOC data set with VOC data set after the city image marks out rubbish region, then build the deep learning platform based on deep learning technique, select the pre-training model on the deep learning platform, carry out the rubbish detection to newly obtaining the city image through deep learning platform and pre-training model after carrying out prior parameter setting to the pre-training model, give the testing result automatically, do not need the special messenger to take the detection of vehicle and shoot registration and artifical rubbish region, can accomplish all-weather city unordered rubbish situation monitoring and detection of abandoning, and is with low costs, consuming time weak point, this detection and the clearance of unordered rubbish in the great convenience city, guarantee urban sanitation and image.
The urban scene garbage detection method and the system select the visual object classification VOC data set as the basic data set for garbage detection, acquire the garbage marked area of the urban image, perform fusion expansion and enrichment of the VOC data set with the VOC data set, then build a deep learning platform based on the deep learning technology, the pre-training model is selected on the deep learning platform, the newly acquired urban image is detected for rubbish through the deep learning platform and the pre-training model after the pre-training model is subjected to prior parameter setting, the detection result is automatically given, no special person is required to take a vehicle for photographing registration and detection of artificial rubbish areas, all-weather monitoring and detection of urban disordered rubbish discarding conditions can be realized, the cost is low, the time consumption is short, therefore, the detection and the cleaning of the disordered discarded garbage in the city are greatly facilitated, and the urban sanitation and image are guaranteed.
The present invention is not limited to the above preferred embodiments, and any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (4)
1. A method for detecting urban scene garbage is characterized by comprising the following steps:
s10: selecting a VOC data set as a basic data set for garbage detection, collecting urban images, selecting the urban images containing disordered discarded garbage as a candidate set, labeling areas of the urban images containing disordered discarded garbage in the candidate set according to a format defined by the VOC data set, and fusing the labeled urban images with existing data in the VOC data set;
s20: building a deep learning platform for garbage detection on the basis of the fused VOC data set, acquiring a pre-training model for garbage detection provided by the deep learning platform on the built deep learning platform, and carrying out adaptive prior parameter setting on the pre-training model;
s30: carrying out garbage detection on the newly acquired urban image by adopting a pre-training model, detecting whether garbage exists in the newly acquired urban image and a garbage existing area, and giving a detection result;
the step S10 specifically includes: collecting urban images, selecting the urban images containing disordered discarded garbage as a candidate set, marking the areas containing the disordered discarded garbage of the urban images in the candidate set by adopting a rectangular selection frame according to a format defined by a VOC data set, randomly dividing the marked urban images into a training set, a verification set and a test set after marking is finished, and respectively fusing the newly obtained training set, verification set and test set with the existing training set, verification set and test set in the VOC data set;
the step S20 specifically includes: verifying the detection precision of different prior parameters on the urban image on the fused VOC data set by using a grid searching method; selecting 0.001 as an initial learning rate, 0.0005 as a weight attenuation and 0.9 as an impulse, selecting N candidate regions from each urban image as subsets, and performing loss back propagation to update the weight of the pre-training model;
the step S30 specifically includes:
preprocessing the newly acquired urban image, specifically comprising cutting and scaling the newly acquired urban image and/or carrying out mean value extraction processing;
and inputting the preprocessed urban image into a deep learning neural network to obtain the classification and position regression of the candidate regions in the urban image so as to obtain a detection result.
2. The urban scene garbage detection method according to claim 1, wherein the step S20 specifically is:
and selecting a Caffe deep learning framework to realize a deep learning platform, and using a ZF Model in the Model zoom as a pre-training Model of the garbage detection task.
3. An urban scene garbage detection system, comprising:
the data fusion module is used for selecting a VOC data set as a basic data set for garbage detection, collecting urban images, selecting the urban images containing disordered discarded garbage as a candidate set, labeling the areas of the urban images containing disordered discarded garbage in the candidate set according to a format defined by the VOC data set, and fusing the labeled urban images with existing data in the VOC data set;
the deep learning platform building module is used for building a deep learning platform for garbage detection on the basis of the fused VOC data set, acquiring a pre-training model for garbage detection provided by the deep learning platform on the built deep learning platform, and carrying out adaptive prior parameter setting on the pre-training model;
the urban image garbage detection module is used for detecting garbage of the newly acquired urban image by adopting a pre-training model, detecting whether garbage exists in the newly acquired urban image and detecting a garbage existing area, and giving a detection result;
the deep learning platform building module verifies the detection precision of different prior parameters on the urban image on the fused VOC data set by using a grid searching method; selecting 0.001 as an initial learning rate, 0.0005 as a weight attenuation and 0.9 as an impulse, selecting N candidate regions from each urban image as subsets, and performing loss back propagation to update the weight of the pre-training model;
the data fusion module collects the urban images, selects the urban images containing disordered discarded garbage as a candidate set, labels the areas containing disordered discarded garbage of the urban images in the candidate set by adopting a rectangular selection frame according to a format defined by the VOC data set, randomly divides the labeled urban images into a training set, a verification set and a test set after the labeling is finished, and fuses the newly acquired training set, verification set and test set with the existing training set, verification set and test set in the VOC data set respectively;
the urban image garbage detection module is used for preprocessing a newly acquired urban image, and specifically comprises the steps of cutting and scaling the newly acquired urban image and/or carrying out mean value extraction processing; and inputting the preprocessed urban image into a deep learning neural network to obtain the classification and position regression of the candidate regions in the urban image so as to obtain a detection result.
4. The urban scene garbage detection system according to claim 3, wherein the deep learning platform building module selects a Caffe deep learning framework to realize the deep learning platform, and uses a ZF model in ModlZoo as a pre-training model of a garbage detection task.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610529468.9A CN106203498B (en) | 2016-07-07 | 2016-07-07 | Urban scene garbage detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610529468.9A CN106203498B (en) | 2016-07-07 | 2016-07-07 | Urban scene garbage detection method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106203498A CN106203498A (en) | 2016-12-07 |
| CN106203498B true CN106203498B (en) | 2019-12-24 |
Family
ID=57465481
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610529468.9A Active CN106203498B (en) | 2016-07-07 | 2016-07-07 | Urban scene garbage detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106203498B (en) |
Families Citing this family (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110119662A (en) * | 2018-03-29 | 2019-08-13 | 王胜春 | A kind of rubbish category identification system based on deep learning |
| CN109165582B (en) * | 2018-08-09 | 2021-09-24 | 河海大学 | Urban street garbage detection and cleanliness assessment method |
| CN108932510A (en) * | 2018-08-20 | 2018-12-04 | 贵州宜行智通科技有限公司 | A kind of rubbish detection method and device |
| US11881019B2 (en) | 2018-09-20 | 2024-01-23 | Cortexia Sa | Method and device for tracking and exploiting at least one environmental parameter |
| CN109389161B (en) * | 2018-09-28 | 2021-08-31 | 广州大学 | Garbage identification evolutionary learning method, device, system and medium based on deep learning |
| CN109376788A (en) * | 2018-10-31 | 2019-02-22 | 重庆爱思网安信息技术有限公司 | A kind of image analysis method based on the high discrimination of deep learning |
| CN109598303A (en) * | 2018-12-03 | 2019-04-09 | 江西洪都航空工业集团有限责任公司 | A kind of rubbish detection method based on City scenarios |
| CN110598784B (en) * | 2019-09-11 | 2020-06-02 | 北京建筑大学 | Machine learning-based construction waste classification method and device |
| CN110659622A (en) * | 2019-09-27 | 2020-01-07 | 北京文安智能技术股份有限公司 | Detection method, device and system for garbage dumping |
| CN110956104A (en) * | 2019-11-20 | 2020-04-03 | 河南华衍智能科技有限公司 | Method, device and system for detecting overflow of garbage can |
| CN111123940A (en) * | 2019-12-27 | 2020-05-08 | 科大讯飞股份有限公司 | Sweeping planning method of sweeping robot, sweeping robot and sweeping system |
| CN112560755B (en) * | 2020-12-24 | 2022-08-19 | 中再云图技术有限公司 | Target detection method for identifying urban exposed garbage |
| CN114758127A (en) * | 2022-04-08 | 2022-07-15 | 山东梧桐城市规划技术服务有限公司 | Urban scene garbage detection system based on big data |
| CN116189099B (en) * | 2023-04-25 | 2023-10-10 | 南京华苏科技有限公司 | Method for detecting and stacking exposed garbage based on improved yolov8 |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103366250A (en) * | 2013-07-12 | 2013-10-23 | 中国科学院深圳先进技术研究院 | City appearance environment detection method and system based on three-dimensional live-action data |
| CN105446183A (en) * | 2015-12-07 | 2016-03-30 | 陆宁远 | Smart city recovery terminal and control method thereof |
| CN105512666A (en) * | 2015-12-16 | 2016-04-20 | 天津天地伟业数码科技有限公司 | River garbage identification method based on videos |
| CN105630882A (en) * | 2015-12-18 | 2016-06-01 | 哈尔滨工业大学深圳研究生院 | Remote sensing data deep learning based offshore pollutant identifying and tracking method |
| CN105772407A (en) * | 2016-01-26 | 2016-07-20 | 耿春茂 | Waste classification robot based on image recognition technology |
-
2016
- 2016-07-07 CN CN201610529468.9A patent/CN106203498B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103366250A (en) * | 2013-07-12 | 2013-10-23 | 中国科学院深圳先进技术研究院 | City appearance environment detection method and system based on three-dimensional live-action data |
| CN105446183A (en) * | 2015-12-07 | 2016-03-30 | 陆宁远 | Smart city recovery terminal and control method thereof |
| CN105512666A (en) * | 2015-12-16 | 2016-04-20 | 天津天地伟业数码科技有限公司 | River garbage identification method based on videos |
| CN105630882A (en) * | 2015-12-18 | 2016-06-01 | 哈尔滨工业大学深圳研究生院 | Remote sensing data deep learning based offshore pollutant identifying and tracking method |
| CN105772407A (en) * | 2016-01-26 | 2016-07-20 | 耿春茂 | Waste classification robot based on image recognition technology |
Non-Patent Citations (3)
| Title |
|---|
| "Faster-RCNN+ZF用自己的数据集训练模型（Matlab版本）";小咸鱼_;《https://blog.csdn.net/sinat_30071459/article/details/50546891》;20160120;第1-6页 * |
| "将数据集做成VOC2007格式用于Faster-RCNN训练";小咸鱼_;《https://blog.csdn.net/sinat_30071459/article/details/50723212》;20160223;第1-4页 * |
| "行人重现检测研究";李千;《万方硕士学位论文》;20160504;第42-46页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106203498A (en) | 2016-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106203498B (en) | Urban scene garbage detection method and system | |
| CN110348445B (en) | Instance segmentation method fusing void convolution and edge information | |
| CN106920229B (en) | Automatic detection method and system for image fuzzy area | |
| CN106960195B (en) | Crowd counting method and device based on deep learning | |
| CN103324937B (en) | The method and apparatus of label target | |
| Morishita et al. | SakuraSensor: Quasi-realtime cherry-lined roads detection through participatory video sensing by cars | |
| CN104615986B (en) | The method that pedestrian detection is carried out to the video image of scene changes using multi-detector | |
| CN108154134A (en) | Internet live streaming pornographic image detection method based on depth convolutional neural networks | |
| CN110781711A (en) | Target object identification method and device, electronic equipment and storage medium | |
| CN107194396A (en) | Method for early warning is recognized based on the specific architecture against regulations in land resources video monitoring system | |
| CN112883921A (en) | Garbage can overflow detection model training method and garbage can overflow detection method | |
| CN113822247A (en) | Method and system for identifying illegal building based on aerial image | |
| CN111898581A (en) | Animal detection method, device, electronic equipment and readable storage medium | |
| CN108769598A (en) | Across the camera video method for concentration identified again based on pedestrian | |
| CN112686172A (en) | Method and device for detecting foreign matters on airport runway and storage medium | |
| CN106339657A (en) | Straw incineration monitoring method and device based on monitoring video | |
| CN111428589B (en) | Gradual transition identification method and system | |
| CN114445780A (en) | Detection method and device for bare soil covering, and training method and device for recognition model | |
| CN113658121A (en) | Line defect detection method, system and computer storage medium | |
| CN109815936A (en) | A kind of target object analysis method and device, computer equipment and storage medium | |
| CN113450573A (en) | Traffic monitoring method and traffic monitoring system based on unmanned aerial vehicle image recognition | |
| CN109165636A (en) | A kind of sparse recognition methods of Rare Birds based on component-level multiple features fusion | |
| CN114078319A (en) | Method and device for detecting potential hazard site of traffic accident | |
| CN115719475A (en) | Three-stage trackside equipment fault automatic detection method based on deep learning | |
| CN118366162B (en) | Image segmentation method and system based on deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |